A cellular video

What if we could watch a cell and all its processes live

This is an idea I have thought of many, many times. Infact, it was an answer I once naively wrote while filling out a job application at a cell imaging company.

I vividly remember picking up Molecular Biology of the Cell in 2007/2008, during my first year as a biotech undergrad, and making detailed notes from it, because I was fascinated. It was immediately clear that there was an astounding level of detail in every cellular process and I had no fair estimate for how many of them were happening at any given second. I remember thinking, it would be so much easier to understand the big picture (and remember), if I could just watch this like a live video in action.

More recently (this week), I read an article called the Virtual Cell by Elliot Hershberg that triggered this memory where he mentioned (and I quote) the “fancy graphical rendering of a cell in its gyrating three-dimensional glory”

So what would happen if we could see inside a human cell and is there value in observing? Would we be able to identify unexpected, upstream or more precise layers of regulation or dysfunction? Or would perturbation experiments be still needed to understand cause and effect. This article is purely me imagining what might be possible one day and what it would lead to.

Imagine: At a transcriptionally active locus, RNA polymerase slides along the DNA, drawing out a ribbon of RNA. Ribonucleotides are added one by one and a readily readable sentence emerges at the base-pair level. In the cytoplasm, tRNAs shuttle in with amino acids and if we zoomed in enough, each residue could be identified by its shape and charge. As they’re linked together, the polypeptide starts curling and twisting until a compact globular protein takes shape, of which the surface residues are also readable.

Pretty cool right, but is there value in observing? I think so. Not in the naive sense that viewing alone would reveal everything that's going on, but rather that this observation would be the richest, most unbiased source of biological data. If we could generate continuous observations at this level of detail across cellular components, contexts and time, then we could train self-learning models that would recognize known rules of biology, but more importantly, reveal new principles: previously uncharacterized regulatory motifs, new mechanisms of molecular coordination, or context-specific variants of canonical pathways. Given enough observations across different cell types, lineages, and states, I believe molecular subtyping and disease detection would look very different

What about target (and thus drug) discovery though? Compound screening would still be needed but our (currently hypothetical but amazing) self-learning model could pinpoint exact mechanisms of action. Drug repurposing would be more systematic and building a virtual cell ie. an in-silico model would be far easier.

Ofcourse all of the above is not feasible in the near future but I do believe improved live cell imaging and using that as data modality should be incorporated along with perturbation experiments in building a virtual cell.

I would love to hear what comes to your mind when you hear “building a cellular video” and if anyone is working on this idea, it would be fantastic to learn more.

Aditi

ps. the visual is generated using https://deepai.org/machine-learning-model/text2img using the prompt “human cell, a movie”